Age and sex estimation using post-mortem CT scout view

PURPOSE: Post-mortem CT (PMCT) is valuable in mass fatality incidents for identification purposes. The aim of our study is to develop and evaluate the performance of “iml-scout”, a Deep Learning software for estimating age and sex from PMCT scout views. MATERIALS AND METHODS: This retrospective study included consecutive decedents with known age and sex who underwent PMCT from January 2015 to January 2024 (derivation cohort) and February to September 2024 (evaluation cohort). We trained “iml-scout” on 60% of the derivation cohort (training set) to estimate age and sex based on frontal and lateral scout views. Age confidence intervals (95%CI) were derived using quantile regression on the validation set (20% of derivation cohort). In the test set (20% of derivation cohort) and the evaluation cohort, we measured mean absolute error (MAE), concordance correlation coefficient (CCC), percentage of age estimations within the 95% CI, sex classification accuracy, and inference duration. RESULTS: The derivation (n = 3008) and evaluation (n = 321) cohorts had median ages of 57 [IQR: 39–71] and 54 [34–70] years, respectively. In the test set (n = 602) and the evaluation cohort, MAE was 6.31 years (95%CI: 5.87–6.75) and 6.57 years (95%CI: 5.96–7.18), with CCC of 0.92 (95%CI: 0.90–0.93) in both. Age estimations were within the 95%CI for respectively 575/602 (95.5%) and 306/321 (95.3%) of cases. Sex classification accuracy was respectively 566/602 (94.0%) and 302/321 (94.1%). Mean inference time was 12.6 ± 0.87 s. CONCLUSION: The “iml-scout” software enables fast, accurate estimation of age and sex from PMCT scout views.